（1）主 题：Application of Compressed Feature Extraction for Electromagnetic Transducer Testing
Electromagnetic Acoustic Transducer (EMAT) has been widely used in industrial Nondestructive Testing (NDT) area. According to the Shannon-Nyquist sampling theorem, an enormous amount of data will be generated during the process of EMAT testing, which increases the complexity of data acquisition system significantly, and makes the real-time signal processing impossible. To solve this problem, we proposed a compressed feature extraction framework based on compressed sensing and convolutional neural networks. The ultimate purpose is to provide a compressed sampling mechanism to EMAT testing system in order to minimize the requisite number of measurements as well as capture useful features from low-dimensional ultrasonic echo signal directly. The proposed framework were be examined through a process of thickness measurement of 813-X70 pipeline, theoretical analysis and results of simulations will be presented to demonstrate its effectiveness and efﬁciency.
Y. Yan received the B.Eng degree in Electronics Engineering from the Xi'an University of Technology, Xi’an, China, M.Sc degree in Mobile and Pervasive Computing from the Newcastle University, Newcastle Upon Tyne, United Kingdom, and the Ph.D degree in Wireless Sensor Networks from the University of York, York, United Kingdom. He is currently a Postdoctoral Research Fellow at the School of Automation Engineering at the University of Electronic Science and Technology of China (UESTC), Chengdu, China. His major research interests include Structure Health Monitoring, Wireless Sensor Networks, Medium Access Control, Deep Reinforcement Learning, and Signal Processing.
（1）主 题：Application of Ground Penetrating Radar in Non-destructive Test on Building Materials
In this talk, we would like to introduce you an efficient non-destructive testing method and to discuss its potential use in testing some materials on civil engineering structures. There are three issues: (1) the introduction of ground penetrating radar (GPR), including its theory, history and applications, (2) the application of GPR to monitor water transfers in building materials, (3) the application of GPR in the structural health monitoring of high-speed rail ballastless undertrack structure.
First, we tell you the story of GPR. GPR is based on one or more ultra-wideband antennas to propagate high frequency electromagnetic (EM) waves (30 MHz – 3 GHz) underground and receive the reflected wave back to the ground. According to the received EM wave’s waveform, amplitude and time variation, we are able to profile the location, structure, and buried depth of the underground medium. It is regarded as one of the most popular EM non-destructive methods. The first GPR meeting was held in Tifton, Georgia, USA in 1986. That is to say, GPR has been studied and discussed for over 30 years. GPR has systematically progressed forward from “Locating and Testing” to “Imaging and Diagnosis” with the Holy Grail of “Seeing the unseen” becoming a reality. Here we will talk about its application civil engineering, generally on two kinds of materials: limestone and concrete.
Then, we propose to analyse the water gradients due to water transfer in limestone/concrete slabs using two non-destructive testing methods with the application of Ground Penetrating Radar (GPR). One uses the two-dimensional Fourier Transform and waveguide model to extract the geometric dispersion of GPR waves. The other one uses the closed-form forward electromagnetic model describing the near-field radar data for two-layer media. Both methods have the same objectivc: to determine the water penetration depth inside the material. The experimental validations have been done by comparing to a reference method: Gammadensimetry.
Finally, our current project “Multi-layer imaging of high-speed railway ballastless track using ground penetrating radar” has some results to present. The three-layer structure, including track slab, CA mortar layer and support plate, of the undertrack has been modelled by FDTD method. We have done a parametric study to see the capability of GPR to see the cracks/ gaps inside the structure with the distribution of steel bars in the track slab.
Xiaoting Xiao received her Ph.D. degree in Electronic, Microelectronic, Nanoelectronic and Microwaves from Nantes University & French Institute of Science and Technology for Transport, Development and Networks, France in 2016. She currently works as a Postdoctoral Fellow in the School of Automation Engineering of UESTC. Dr. Xiao is working in the field of Electromagnetic Non-destructive Testing. Her interests include the electromagnetic modeling, imaging algorithms for ground penetrating radar and the inversion methods for high resolution non-destructive characterization of civil engineering materials.
（1）主 题：Tobit Kalman Filtering with Incomplete Information
In this talk, we briefly discuss Tobit Kalman filtering problem for a class of discrete time-varying systems with incomplete information. The incomplete information includes: (1) fading measurements; (2) transmission failures; (3) censored measurements. Censored measurements arise frequently in engineering practice involving a large number of low-cost off-the-shelf sensors.
First, the censored measurements are described by the Tobit measurement model and the fading measurements are characterized by the Lth-order Rice fading channel model capable of accounting for not only the packet dropout but also the communication phenomenon. The measurement fading occurs in a random way where the fading probabilities are regulated by a set of mutually independent Gaussian random variables. By resorting to the state augmentation technique and the orthogonality projection principle, the Tobit Kalman filter (TKF) is designed in the presence of fading measurements. In the course of filter design, several state-augmentation-induced terms are introduced, all of which can be calculated recursively or off-line.
On the other hand, sensor measurements often suffer from intermittent failures in data transmissions, and an effective way to improve the transmission reliability is to adopt the redundant channel transmission protocol. The Tobit Kalman filtering problem is investigated for linear discrete time-varying systems with censored measurements, intermittent failures and time-correlated multiplicative measurement noises under the redundant channel transmission protocol. The Tobit regression model is first modified to take into account the complexities contributed by measurement noises, intermittent failures as well as the redundant channels. Then, an optimal Tobit Kalman filter is designed based on the modified Tobit regression model. In the developed algorithm for the filter design, several new terms are introduced to reflect addressed the complexities, all of which can be calculated recursively or off-line.
Finally, we explore oscillator example about tracking ballistic rate to demonstrate the effectiveness of the proposed filtering algorithms and the filtering performance.
H. Geng received the Bachelor degree from Civil Aviation Flight University of China and Ph.D. degree from the Northwestern Polytechnical University in 2011 and 2017, respectively. He currently works as a Postdoctoral Fellow in the School of Automation Engineering of UESTC. Dr. Geng is working in the field of complex systems modeling, estimation and fusion. His main research interests include fault diagnosis, information fusion, signal processing and target tracking.
编辑：罗莎 / 审核：林坤 / 发布者：林坤